Based on the excellent responses here, the conclusion is that this guy is completely wrong and that query plans are definitely useful and even required. Now I think it is interesting to figure out how it can be that he is so wrong. It seems to me the simplest explanation is that the pg devs aren't eating enough of their own dogfood and are just unaware of a certain uses their product is put to? However, I would expect them to have been supplying consultancy services to exactly the users that come across these kinds of problems. In that case, the 'no query plans' must also have become stuck in their mindset, to remain unmodified after coming across practical uses cases that actually require them.